Re: [PATCH] maint: avoid reading lots of data from /dev/urandom

[Jim Meyering]

Pádraig Brady wrote:
> On Solaris 10 at least you need to use dd iflag=fullblock
> to avoid short reads.  Also it's relatively slow when
> you just want non NUL data as opposed to pseudo random data.
>
> * cp/fiemap-empty: Convert /dev/zero rather than using /dev/urandom.
> * dd/sparse: Likewise.

Thanks.  Do either of these avoid an actual failure?

With those changes, the affected files are then very compressible.
With some file system types/options, won't they occupy far fewer
blocks than before?

>  tests/cp/fiemap-empty |    2 +-
>  tests/dd/sparse       |    4 ++--
>  2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/tests/cp/fiemap-empty b/tests/cp/fiemap-empty
> index 4ba4ee3..81926f3 100755
> --- a/tests/cp/fiemap-empty
> +++ b/tests/cp/fiemap-empty
> @@ -64,7 +64,7 @@ rm empty.big cp.test
>  # Note until we support fallocate, we will not maintain
>  # the file allocation.  FIXME: amend this test when fallocate is supported.
>  fallocate -l 10MiB -n unwritten.withdata || framework_failure_
> -dd count=10 if=/dev/urandom conv=notrunc iflag=fullblock 
> of=unwritten.withdata
> +dd count=10 if=/dev/zero | tr '\0' '1' | dd conv=notrunc 
> of=unwritten.withdata
>  cp unwritten.withdata cp.test || fail=1
>  test $(stat -c %s unwritten.withdata) = $(stat -c %s cp.test) || fail=1
>  cmp unwritten.withdata cp.test || fail=1
> diff --git a/tests/dd/sparse b/tests/dd/sparse
> index 35ddda9..f132924 100755
> --- a/tests/dd/sparse
> +++ b/tests/dd/sparse
> @@ -45,8 +45,8 @@ cmp file.in file.out || fail=1
>  # Setup for block size tests: create a 3MiB file with a 1MiB
>  # stretch of NUL bytes in the middle.
>  rm -f file.in
> -dd if=/dev/urandom of=file.in bs=1M count=3 iflag=fullblock || fail=1
> -dd if=/dev/zero    of=file.in bs=1M count=1 seek=1 conv=notrunc || fail=1
> +dd if=/dev/zero bs=1M count=3 | tr '\0' '1' | dd of=file.in || fail=1
> +dd if=/dev/zero of=file.in bs=1M count=1 seek=1 conv=notrunc || fail=1
>
>  kb_alloc() { du -k "$1"|cut -f1; }